feat(processing): 快照流程锁内复核收敛 + 异常显式清除语义 + 语义矩阵定稿 (ACM2-29 P2-5/M3)

- SnapshotFlow 收敛到 v2 §5 事务顺序：锁外仅解析校验；取得 PIPELINE_LOCK 后
  锁内复核快照身份（last_message_id 重放短路）→ 锁内读代与当前航班态 →
  计算 nextState → 写入 → CAS → 事件 → SUCCEEDED。锁内 CAS 失败属数据异常，
  一律回滚 FAILED(INFRA)，删除「版本号相同即视为重放」的推断路径（评审 F7/F8 收口）
- M3/F4：异常对象（FDIV/FRET/FLAB）显式清除语义——null 字面量/空对象/空串
  翻译为 Clear 命令；FlightNextState.clearedKeys 携带本次清除键；增量路径按
  映射把前缀标量列置 NULL；库内清除后读视图无该键（线格式保持键缺失，显式
  清除表达留待阶段 2 按下游契约定）
- StubFlightSchd 增量分支改为 nextState 权威整体替换（引擎已合并当前态），
  天然承载 Clear；flattenScalarsOnly 去除冗余循环（L2）
- docs/flight-state-semantics.md：16 类结构语义矩阵定稿（P0-A 交付物落库）
- 新测试：引擎 0 标记/异常清除/混排 fail-fast 单测；JDBC 异常清除回环用例

验证：MSGX_PG_PORT=5433 真实 PG ./gradlew test --rerun-tasks
105 用例 0 失败 0 跳过

src/main/kotlin/com/gzzn/omms/msgexchange/infra/persistence/jdbc/JdbcPgRepositories.kt

@@ -518,8 +518,6 @@ class JdbcFlightSchdRepository(
 
     /** v2：仅标量/异常/文本列 flatten；集合键由明细表读写，不经过槽位列。 */
     private fun flattenScalarsOnly(scalars: Map<String, String>): Map<String, String?> {
-        val fields = linkedMapOf<String, String>()
-        scalars.forEach { (key, value) -> fields[key] = value }
         val out = mutableMapOf<String, String?>()
         scalars.forEach { (key, value) ->
             when {
@@ -527,7 +525,7 @@ class JdbcFlightSchdRepository(
                 key in setOf("SRVT", "VIPF", "MAFL") -> out["${key}_TEXT"] = value
             }
         }
-        flattenExceptions(fields, out)
+        flattenExceptions(scalars, out)
         return out
     }
 
@@ -936,15 +934,31 @@ class JdbcFlightSchdRepository(
         }
     }
 
-    /** v2 增量：仅更新出现的标量列 + 追踪字段。 */
+    /** 显式清除键 → 需要置 NULL 的物理列（异常前缀列 / 文本列 / 普通标量列）。 */
+    private fun clearColumnsFor(keys: Set<String>): List<String> = keys.flatMap { key ->
+        when {
+            key == "FDIV" -> listOf("FDIV_DDES", "FDIV_DDIR", "FDIV_REMC")
+            key == "FRET" -> listOf("FRET_REID", "FRET_RSN")
+            key == "FLAB" -> listOf("FLAB_ARES", "FLAB_RSN")
+            key == "SRVT" -> listOf("SRVT_TEXT")
+            key == "VIPF" -> listOf("VIPF_TEXT")
+            key == "MAFL" -> listOf("MAFL_TEXT")
+            key in SCALAR_KEY_SET -> listOf(key)
+            else -> emptyList()
+        }
+    }
+
+    /** v2 增量：仅更新出现的标量列 + 显式清除列置 NULL + 追踪字段。 */
     private fun upsertIncrementalScalars(state: com.gzzn.omms.msgexchange.domain.flight.FlightNextState, now: Instant) {
         val stored = flattenScalarsOnly(state.scalars)
-        if (stored.isEmpty()) {
+        val clearedColumns = clearColumnsFor(state.clearedKeys)
+        if (stored.isEmpty() && clearedColumns.isEmpty()) {
             updateTrackingFields(state, now)
             return
         }
-        val assignments = stored.keys.map { "$it = ?" }.toMutableList()
-        assignments += listOf("last_message_id = ?", "state_version = ?", "updated_at = ?")
+        val assignments = stored.keys.map { "$it = ?" } +
+            clearedColumns.map { "$it = NULL" } +
+            listOf("last_message_id = ?", "state_version = ?", "updated_at = ?")
         val sql = "UPDATE flight_schd SET ${assignments.joinToString(", ")} WHERE flid = ?"
         val sqlTimestamp = now.toSqlTimestamp()
         ds.update(sql) { ps ->

src/main/kotlin/com/gzzn/omms/msgexchange/infra/stub/StubRepositories.kt

@@ -335,21 +335,17 @@ class StubFlightSchd : FlightSchdRepository {
                 )
                 detailCollections[state.flid] = state.collections
             } else {
-                val mergedScalars = (existing?.fields?.filterKeys {
-                    it == "FLID" || it !in com.gzzn.omms.msgexchange.infra.persistence.jdbc.FlightDetailTables.DETAIL_COLLECTION_KEYS
-                } ?: emptyMap()) + scalarFields
+                // nextState 由引擎在当前态上合并而来，是完整权威态：直接整体替换（含清除语义）
                 records[state.flid] = Record(
                     flid = state.flid,
                     fday = existing?.fday,
-                    fields = mergedScalars,
+                    fields = scalarFields,
                     stateVersion = state.stateVersion,
                     lastMessageId = state.lastMessageId,
                     createdAt = existing?.createdAt ?: now,
                     updatedAt = now,
                 )
-                val mergedCollections = (detailCollections[state.flid] ?: emptyMap()).toMutableMap()
-                state.collections.forEach { (key, items) -> mergedCollections[key] = items }
-                detailCollections[state.flid] = mergedCollections
+                detailCollections[state.flid] = state.collections
             }
         }
     }